AI implementation is a winner-take-all race, analyst says

Conventional wisdom states that in the early stages of AI adoption, enterprises should grab low-hanging fruit. They should start a point project to automate a repeatable process and pocket the efficiency gains. "AI is more about growth than efficiency," said Bughin, director of the McKinsey Global Institute at McKinsey and Company, in a webinar hosted by MIT. "Don't look for marginal projects. According to Bughin, technologies like natural language processing and generation, computer vision, and advanced robotics -- some of the most advanced areas of AI today -- don't generally have much to do with efficiency or the automation of existing processes. They have the potential to enable entirely new processes, making enterprises more competitive in existing markets, helping them reach into new markets and enabling them to potentially lead the development of new products. He mentioned Uber as an example of a company that used machine learning to create a market. Bughin's advice runs counter to what a lot of enterprises are doing with AI today. Most are taking a cautious approach and only proceeding with AI implementations in incremental ways, if at all. At the Gartner Data and Analytics Conference in Dallas in March, analyst Whit Andrews said most enterprises today assume that they are behind their competition on implementing AI, but the truth is that a relatively small number of companies are using AI, and even fewer in a pervasive manner throughout their organization. IBM chief architect Ruchir Puri explains the top three challenges to implementing AI in business. But Bughin said businesses shouldn't aim to stay even with their competition; they should look to beat them to new customers and markets. And since AI implementation is relatively limited at this point, making smart use of it can be a substantial competitive advantage. "There's no value in waiting to get to AI," he said. "Your competitors will, so you will either implement it or be playing catch up.